Skip to main content

pg_mimic

A pure-Python, asyncio PostgreSQL server protocol emulator — the Postgres analog of mysql-mimic. Embed it in your own process, subclass Session to answer queries with arbitrary Python logic, and get a fully wire-compatible server that real Postgres clients, drivers, ORMs, and tools (psql, psycopg, asyncpg, JDBC, ...) can connect to.

Useful for testing, proxies/virtual databases, query interception, and fronting non-Postgres backends with a Postgres-speaking API.

Install

pip install pg-mimic

Quick start

import asyncio
from pg_mimic import PgServer, ResultColumn, Session


class MySession(Session):
    async def describe(self, sql, param_oids):
        return [ResultColumn.for_type("a", str), ResultColumn.for_type("b", int)]

    async def query(self, sql, params):
        yield ("hello", 1)
        yield ("world", 2)


async def main():
    server = PgServer(session_factory=MySession)
    await server.start_server(host="127.0.0.1", port=5432)
    await server.serve_forever()


asyncio.run(main())
$ psql "host=127.0.0.1 port=5432 user=test dbname=test" -c "select * from anything"
   a   | b
-------+---
 hello | 1
 world | 2
(2 rows)

More examples: examples/simple.py, examples/echo.py (logs every statement it receives -- good for poking at the server interactively with psql and watching what comes through), examples/parameterized.py (real bind parameters), examples/dbapi_proxy.py (fronting a real sqlite3 database).

The Session API

Override two methods:

  • describe(sql, param_oids) -> list[ResultColumn] | None — declares column shape (names + Postgres type OIDs). Return None for statements that produce no rows (INSERT/UPDATE/DELETE without RETURNING, etc). Column shape is always a declared fact, never inferred by peeking at row data.
  • query(sql, params) -> AsyncIterator[tuple] — the actual rows, as an async generator (or a coroutine returning any iterable). params are the real out-of-band bind parameter values as decoded text strings (not textually interpolated into sql) — pg_mimic decodes them itself even when a client sends them in binary format (psycopg, JDBC, and most other drivers do this by default for common scalar types).

Optionally override schema() to describe your tables for information_schema introspection (see below).

Why two methods instead of one

Postgres's extended query protocol (Parse/Bind/Describe/Execute/Sync) genuinely separates "what columns will this return" from "run it and give me rows" — a client can ask Describe before ever issuing Execute. describe() answers that without touching any row source; query() is only ever invoked once Execute actually runs, so side effects happen exactly when a real Postgres server would trigger them, not earlier.

Real parameterized queries and prepared statements

Both the simple query protocol (psql, one-shot queries) and the extended protocol (parameterized queries, prepared statement reuse) are fully supported and drive the exact same Statement/Portal machinery internally — a prepared statement Parsed once can be Bind+Executed repeatedly with different parameters, and large results stream incrementally (Execute's row-limit / PortalSuspended semantics) rather than being forced into memory all at once, as long as query() yields lazily.

Power tier: full Statement/Portal control

For sessions that front a real backing connection (see examples/dbapi_proxy.py), override prepare() directly instead of describe()/query():

from pg_mimic import Statement, Portal


class MyStatement(Statement):
    async def describe(self): ...
    def bind(self, params) -> Portal: ...


class MySession(Session):
    async def prepare(self, sql, param_oids) -> Statement:
        return MyStatement(...)

This maps 1:1 onto a real database driver's own prepare/bind/execute stages with no translation overhead.

Authentication

Four pluggable mechanisms, matching real Postgres's pg_hba.conf methods:

from pg_mimic import PgServer, TrustAuthPlugin, Md5PasswordAuthPlugin, ScramSha256AuthPlugin, SimpleIdentityProvider

# trust (default): any username/password accepted
server = PgServer(session_factory=MySession)

# password-protected
server = PgServer(
    session_factory=MySession,
    auth_plugin_factory=lambda username: ScramSha256AuthPlugin(),  # or Md5PasswordAuthPlugin / ClearTextPasswordAuthPlugin
    identity_provider=SimpleIdentityProvider({"alice": "s3cret"}),
)

What's handled automatically

A small middleware layer (mirroring mysql-mimic's design) intercepts common statement shapes before they ever reach your Session, so real clients/ORMs/psql work smoothly without you having to handle them:

  • Transaction control: BEGIN/START TRANSACTION, COMMIT/END, ROLLBACK — including the failed-transaction state (25P02) real Postgres enters after an error until ROLLBACK.
  • SET/SHOW (session variables).
  • Static, no-table SELECTs: SELECT 1, SELECT version(), SELECT current_database(), etc.
  • information_schema.tables / information_schema.columns, built from your Session.schema().
  • Multi-statement simple-query batches ("BEGIN; INSERT ...; COMMIT;" sent as one 'Q' message, e.g. by psql -f script.sql) are split into individual statements (via sqlglot), each getting its own RowDescription/DataRow*/CommandComplete — not silently merged or truncated.

Anything else falls through to your describe()/query() (or prepare()).

Known v1 limitations

  • Text format only. Both parameters and results are text-format on the wire (matching pg8000's own client-side posture — it never requests binary either). Binary parameter values are decoded transparently for common scalar types (ints, floats, bool, bytea, date/timestamp, uuid) since several real clients send these in binary by default; binary result format requests get a clear error rather than silently-wrong bytes.
  • No TLS. SSLRequest is answered correctly ('N', continue in plaintext) so opportunistic clients still connect, but there's no ssl.SSLContext upgrade path yet.
  • pg_catalog is not emulated (only information_schema) — tools that specifically query pg_catalog.pg_class etc. won't see your schema.
  • Describe(Statement) before any Bind can only answer accurately for statement shapes pg_mimic (or your describe()) can determine without real parameter values — inherent to being a planner-less mimic, not something a workaround can fully close. In practice this rarely matters: most drivers (including psycopg's default flow) Describe the portal, after Bind, which is always exact.

Development

pip install -e ".[dev]"
pytest

Metadata

Release files for pg-mimic 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for pg-mimic 0.1.0
File Size Uploaded
pg_mimic-0.1.0.tar.gz 33.4 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for pg-mimic 0.1.0
File Interpreter ABI Platform
pg_mimic-0.1.0-py3-none-any.whl Python 3 none any Details

Total release size: 60.9 kB

Release files / pg_mimic-0.1.0.tar.gz

Download URL pg_mimic-0.1.0.tar.gz
Size 33.4 kB
Tags Source
SHA-256 checksum
How to use checksums
c010de074f1bbfc16832983276febcecd794d3d562d7971494572c97f1031af8
BLAKE2b-256 checksum
How to use checksums
8bc61e2486e70a1d1f45488af0610124764a356ca073c8c61419273cf3ba3152
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Aug 11, 2026.

Transparency log

Release files / pg_mimic-0.1.0-py3-none-any.whl

Download URL pg_mimic-0.1.0-py3-none-any.whl
Size 27.5 kB
Tags Python 3
SHA-256 checksum
How to use checksums
505e4db156228f61cfca330e8ae78d9831df8d75a32ea8bad4209ddcd3b05262
BLAKE2b-256 checksum
How to use checksums
6a72476a6de43fe9435c2579b4f26a366ae1b0e1b0b87e72e33bc3e5b413d2bc
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Aug 11, 2026.

Transparency log

Release history Release notifications | RSS feed

0.1.3

2 release files

0.1.2

2 release files

0.1.1

2 release files

This release

0.1.0 This release

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page